PATCH: DAGCombiner - Use correct value type when checking legality of BR_CC - Changes to NVPTX, Mips, Hexagon, MBlaze, X86, and XCore targets

[Jyotsna Verma]

Hi Tom,

For Hexagon, the BR_CC node must be expanded for i1,i32, f32 and f64.

Thanks,
Jyotsna

--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by
The Linux Foundation


>-----Original Message-----
>From: llvm-commits-bounces at cs.uiuc.edu [mailto:llvm-commits-
>bounces at cs.uiuc.edu] On Behalf Of Tom Stellard
>Sent: Thursday, February 28, 2013 2:40 PM
>To: Justin Holewinski
>Cc: llvm-commits; Justin Holewinski
>Subject: Re: PATCH: DAGCombiner - Use correct value type when checking
>legality of BR_CC - Changes to NVPTX, Mips, Hexagon, MBlaze, X86, and XCore
>targets
>
>On Thu, Feb 28, 2013 at 09:04:04AM -0500, Justin Holewinski wrote:
>> For NVPTX, the Expand action should be set on all of {i1, i8, i16,
>> i32, i64, f32, f64}
>>
>
>Hi Justin,
>
>I've attached an updated version of the patch with these changes.
>
>-Tom
>
>
>>
>> On Wed, Feb 27, 2013 at 6:14 PM, Tom Stellard <tom at stellard.net> wrote:
>>
>> > Here is an updated patch.  The previous version had some unrelated
>> > changes included in it.
>> >
>> > -Tom
>> >
>> > On Wed, Feb 27, 2013 at 05:21:59PM -0500, Tom Stellard wrote:
>> > > Forgot the patch...
>> > >
>> > > On Wed, Feb 27, 2013 at 05:17:32PM -0500, Tom Stellard wrote:
>> > > > Hi,
>> > > >
>> > > > The attached patch fixes a bug where the DAGCombiner was passing
>> > > > the MVT::Other value type to TLI.isOperationLegalOrCustom() for
>> > > > a BR_CC
>> > node.
>> > > > LegalizeDAG uses the value type of the comparison operands, and
>> > > > I think this is what the DAGCombiner should be using as well.
>> > > >
>> > > > I've updated the *ISelLowering files for the affected targets,
>> > > > but backend code owners should double check my work.
>> > > >
>> > > > Thanks,
>> > > > Tom
>> > > > _______________________________________________
>> > > > llvm-commits mailing list
>> > > > llvm-commits at cs.uiuc.edu
>> > > > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>> >
>> >
>> > _______________________________________________
>> > llvm-commits mailing list
>> > llvm-commits at cs.uiuc.edu
>> > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>> >
>> >
>>
>>
>> --
>>
>> Thanks,
>>
>> Justin Holewinski